>>109460990
>>109439887
Ok I made some progress today. Implemented a simple """neural network""" for xor in go and that kinda works. With all the copy and paste for forward and backward passes.
Today I started to implement some basic matrix operations, they seem to work. Currently structuring the network with them.
le xor output:
user@desktop:goneuronsgo/ [master*] $ go run .
2026/08/05 13:39:07 INFO Hello, Neurons!
Input network:
--------------------------------------------------
L0
or: {w: [0.67646, 0.46139], b: 0.50855}
nand: {w: [0.42979, 0.79780], b: 0.38837}
--------------------------------------------------
L1
and: {w: [0.81928, 0.33818], b: 0.97305}
--------------------------------------------------
0) input: [0 0]
output: 0.029, expected: 0.0
1) input: [0 1]
output: 0.974, expected: 1.0
2) input: [1 0]
output: 0.974, expected: 1.0
3) input: [1 1]
output: 0.027, expected: 0.0
Output network:
--------------------------------------------------
L0
or: {w: [6.22042, 6.27605], b: -2.73054}
nand: {w: [4.26334, 4.27605], b: -6.58155}
--------------------------------------------------
L1
and: {w: [8.77905, -9.51845], b: -4.03387}
--------------------------------------------------
I think this time I can really do it.
Since go has no fancy operator overloading I was also thinking about refreshing my C++ if I got it working in go. But we will see then if C++ or VHDL.
Gj anon
